2. The data
The high-resolution spectra (R=65000) were obtained at ESO La Silla, with the CAT telescope and the CES spectrograph with its long camera, over a period of 6 nights (21-26 April 1996) that was not free of bad weather. On about 2600 useful columns of ESO CCD #38, the spectra sample the wavelength region of the SiIII triplet ( 4552, 4567, 4574 Å, see Fig. 2), and have a S/N ratio typically between 500 and 1000. With the exception of a few spectra, the exposure time was kept shorter than 20 minutes (typically 15 minutes).
The spectra were reduced using standard packages in IRAF running on a Linux PC. Wavelength calibration was done from ThAr calibration spectra. We used the CCD overscan region to determine the bias level. CCD pixel-to-pixel variations were removed by flat fielding with dome flats; dome flats also provided a first rough correction for the continuum shape which is heavily affected by vignetting of the light beam. Bad columns were removed by linear interpolation of pixel intensities in adjacent columns. One-dimensional spectra were extracted after subtracting a global fit to background and scattered light. All spectra were shifted to, and acquisition times were transformed to, the heliocentric frame.
At first we normalized the spectra by fitting a cubic spline with typically 10-15 segments to the regions in between the obvious absorption lines in the extracted spectra. Since the line profiles are rather broad and the instrumental setup is not stable with respect to continuum shape, this normalization technique gives considerable scatter in the equivalent width of the lines. Therefore we employed a different strategy as well: we used a spectrum with very high S/N as a template and divided this into each individual spectrum. Then we `normalized' these quotient spectra by fitting similar functions as mentioned above; it turned out that the continuum in the quotient spectra is much easier to fit than in the raw spectra. Then we constructed, using the fits to the quotient spectra and a normal continuum fit of the template spectrum, a final set of normalized spectra. For this final set the scatter in the equivalent width of the lines is reduced by about 50% with respect to the results of the first normalizing method.
2.1. Line profile variations
During the six observing nights a total of 30 spectra were taken, most of them in the last three nights of the run. In Fig. 2 we present a grey-scale representation of the data of Sco taken on these three nights.
From Figs. 1 and 2 it is obvious that the line-profile variations in Sco closely resemble that of known non-radially pulsating early-type stars (e.g. Oph, Vogt & Penrod 1983; Sco, Smith 1986; Per, Gies & Kullavanijaya 1988; etc.), and that these spectra appear as if they were the result of NRP model calculations (e.g. Kambe & Osaki 1988, Schrijvers et al. 1997, Townsend 1998, see also Sect. 4). Given the spectral type of the star, the observed pulsation frequency (see Sect. 4.2) and the above findings, we conclude that Sco is most probably a new Cephei star (see also Sect. 5).
2.2. Equivalent width and line centroid variations
Fig. 3 shows the measured equivalent width (EW) variations, and the variations of the centroid velocity of the 4552 Å line (computed for 4552.6 Å rest wavelength). The errors on the points are derived from the S/N of the spectra, and do not include an estimate of continuum misplacement. For the equivalent width we suspect that the main part of variability is caused by errors in the normalization of the spectra. The overplotted equivalent width measurements of the 4567 Å line confirm this: there is no relation between the equivalent width variations of the two lines that are of the same triplet.
The line centroid measurements show a change of about 1 km/s between the beginning and the end of the night (see Fig. 3). The other lines of the triplet show similar behaviour. For our spectra the 1 km/s change corresponds to a spectral shift of just less than one CCD pixel, whereas the slit (2.5 arcsec width) was imaged onto approximately four pixels. Given the complicated light path from the telescope to the slit we cannot exclude that the gradual line shifts are the result of an hour-angle dependent observational effect.
The above arguments and the lack of temporal correlation of the variability of these quantities with the main pulsation period of the star (see below), suggest that these EW and centroid variations might well be instrumental.
© European Southern Observatory (ESO) 1998
Online publication: September 30, 1998